中文摘要：

文章为缓解农业遥感农作物分类识别方面的分类精度与成本、空间分辨率与识别方法之间的矛盾问题提供新的方法。以中分辨率的Landsat TM/ETM＋遥感影像为主要数据源,农田系统图斑的矢量数据作为参考,面向对象方法和CART决策树算法等相结合,提出一种面向对象决策树的农作物分类识别方法。文章分别用提出的方法和传统的决策树分类方法对新疆焉耆盆地2011至2014年的各种地物进行分类识别,4年的平均分类精度92.09%（Kappa系数为0.91）。新方法不仅提高农作物分类识别的总精度,还能比较完整地保持农田作物的轮廓。该成果为中等分辨率的遥感影像以高精度的农作物分类识别、信息提取和动态监测提供可能。

英文摘要：

Crop acreage and type information is very important for accurate estimating crop production,the policy-making of agriculture and ensuring national food security.Crop mapping by using the remotely-sensed images provide basic information for further geographical and ecological researches. In the current agricultural production and management,there is an urgent need of the direct and immediate information guide. In order to ease the contradictions between classification precision and cost,spatial resolution of images and recognition method,a new method so called object-oriented based approach of decision tree for crops classification and recognition was proposed using Landsat TM/ ETM＋ remotely sensed images and vectored capital farmland parcels as basic data sources employing object-oriented methodology and diction tree algorithm. The yearly land-use and land cover changes of Yanqi basin in Xinjiang,China between 2011 and 2014 were studied by the new and traditional methods of diction tree respectively,and get an average high overall precision of 92.09 % with the Kappa coefficient of 0.91. The new method not only improved the precision of crops classification and recognition,but also extracted figure outlines of capital farmland parcels perfectly.The new method brought forwarded in this paper provide possibility for crops classification and recognition,extraction of area information and dynamic monitoring with high precision from remotely sensed images of medium spatial resolution.